LiveHow do we verify the human an AI agent escalates to?

AI agent human escalation

When an autonomous workflow hands a decision to a person, verify that a person actually took it — and record that they did.

The human in the loop, proven

An agent escalating a decision needs more than a click from an authenticated session; it needs evidence that a human participated. A presence check bound to the escalated action supplies exactly that.

The sealed record is what later shows the human step was real rather than automated away.

The artifacts behind this page

What this page does not claim

Limits

  • Presence verifies humans today. Verifying an AI agent as a principal in its own right is a direction, not a shipped capability.

One next step

Read the API reference